AI-generated medical data can sidestep usual ethics review, universities say
4 hours ago
- #AI in Medicine
- #Synthetic Data
- #Ethics in Research
- Medical researchers in Canada, the US, and Italy are using AI-generated synthetic data from real patient information without ethics board approval.
- Synthetic data is created by training AI models on real human medical data to produce datasets with similar statistical properties but no real patient information.
- Institutions like IRCCS Humanitas Research Hospital, CHEO, Ottawa Hospital, and WashU Medicine have waived ethics board reviews for synthetic data research.
- Benefits of synthetic data include protecting patient privacy, easier data sharing, and faster research progress.
- Washington University started waiving ethical reviews in 2020, citing that synthetic data doesn't qualify as 'human-subject research' under US federal rules.
- In Italy, Humanitas AI Center uses synthetic data without ethics board consent if original patient data was collected with AI analysis permission.
- Ontario's 2004 Personal Health Information Protection Act allows non-personal data creation without patient consent.
- Canadian hospitals like CHEO and Ottawa Hospital concluded in 2024 that synthetic data doesn't require ethics oversight as it doesn't meet human research definitions.
- Studies accessing real patient data to create synthetic datasets still need ethics approval but often qualify for low-risk consent waivers.